ShapeSearch: A Flexible and Efficient System for Shape-based Exploration of Trendlines
Identifying trendline visualizations with desired patterns is a common and fundamental data exploration task. Existing visual analytics tools offer limited flexibility and expressiveness for such tasks, especially when the pattern of interest is under-specified and approximate, and do not scale well when the pattern searching needs are ad-hoc, as is often the case. We propose ShapeSearch, an efficient and flexible pattern-searching tool, that enables the search for desired patterns via multiple mechanisms: sketch, natural-language, and visual regular expressions. We develop a novel shape querying algebra, with a minimal set of primitives and operators that can express a large number of ShapeSearch queries, and design a natural-language and regex-based parser to automatically parse and translate user queries to the algebra representation. To execute these queries within interactive response times, ShapeSearch uses a fast shape algebra-based execution engine with query-aware optimizations, and perceptually-aware scoring methodologies. We present a thorough evaluation of the system, including a general-purpose user study, a case study involving genomic data analysis, as well as performance experiments, comparing against state-of-the-art time series shape matching approaches---that together demonstrate the usability and scalability of ShapeSearch.